All classes

(GW = 6.986; p = 0.137;df = 4)

46.9% 41.5% 48.8% 53.7% 26.8%

Kendall correlations between the benthic indices and physical-chemical variables indicated significant correlations in the Mondego and Tejo River estuaries, with Shannon- -Wiener, Margalef and EQR results correlated with DIN concentrations (P<0.01) (Table 5.4).

DISCUSSION

Indices of biotic integrity have long been recognized as useful tools to measure biological responses to pollution, identify the need to apply mitigation measures and evaluate the efficiency of those measures, mainly because these indices: (1) are documented to accurately reflect both watershed level stressors and resulting exposure variables (Dauer et al., 2000; Salas et al., 2006), (2) reduce large amounts of data to a meaningful single value, and (3) are valuable tools to communicate complex data to a general audience (Weisberg et al., 1997; Aubry & Elliott, 2006).

Nevertheless, the efficiency of indices depends on a predictive understanding of the responses of the component metrics of the index to environmental stress and disturbance, at multiple spatial and temporal scales (Andersen, 1997), that might vary considerably between the region for which they were developed and other regions where they might be used.

The present study compared two multimetric indices, B-IBI and TICOR approach, in estuaries with different hydromorphological characteristics and different human pressures, namely the Mondego, Tejo and Mira estuaries. B-IBI was developed using a priori specific physical-chemical criteria (i.e., bottom-water dissolved oxygen concentrations, levels of

sediment contaminants, and/or total organic carbon of the sediment) to: (1) identify samples considered undegraded, degraded or indeterminate, (2) select metrics based upon significant differences, in an ecologically meaningful manner, between undegraded and degraded data, and (3) determine thresholds for scoring each metric, using only data from samples declared undegraded. As such, the B-IBI approach relies upon reference or minimally-impacted conditions. In contrast, the TICOR approach is based solely upon best professional judgment to determine thresholds for each of the component indices and the combined index.

Physical-chemical reference conditions

Most studies recently published on the efficiency of benthic indices to separate between degraded and undegraded conditions are based on prior knowledge of pressures acting over different locations (Salas et al., 2004; 2006; Labrune et al., 2006; Quintino et al., 2006) or using physical-chemical indicators (Weisberg et al., 1997; Paul et al., 2001; Marín- -Guirao et al., 2005). The latter approach was used in the present study since no reference conditions were available and the results indicate that none of the estuaries seems to meet the WFD objective of Good physical-chemical status. Eutrophication problems in the Mondego River estuary and contamination by heavy metals in the Tejo River estuary are well documented, but the Mira River estuary has been characterized as a relatively pristine ecosystem (Marques et al., 1993; Carvalho et al., 2005). Nevertheless, physical-chemical elements measured during the present study showed significant concentrations of nutrients and metals, as well as the occurrence of low oxygen levels in some areas of the Mira River estuary. These results confirm the effect of known pollution sources, such as the Odemira sewage discharge, the irrigation perimeter of the Mira catchment area and mining activities that used to occur in upstream areas.

In determining between quality classes using physical-chemical indicators, specific threshold values separating quality classes are assumed to have a meaningful relationship to pollution-induced alterations in benthic populations. While the use of ERM and ERL values is widely accepted as a reference for heavy metals contamination (e.g. Weisberg et al., 1997;

Mucha et al., 2003), criteria for dissolved oxygen and eutrophication descriptors still lack consensus among the scientific community. According to the criteria of the United States National Estuarine Eutrophication Assessment (NEEA) or the OSPAR Comprehensive procedure, oxygen concentrations ranging between 2 mg l^-1 and 5-6 mg l^-1 may cause stress responses in invertebrate fauna (Bricker et al., 2003; OSPAR Commission, 2003). The application of either the NEAA or OSPAR dissolved oxygen criteria to our study would have produced very similar classifications to ours, but using the criteria of Weisberg et al. (1997) would have resulted in all stations classified as undegraded, since dissolved oxygen concentrations were consistently

high along time (unpublished data). On the other hand, the OSPAR Commission criteria for DIN, based on background concentrations, are much less stringent than the NEEA criteria, since concentrations of 44 µmol l^-1 and 34 µmol l^-1were defined for the Mondego and Tejo estuaries, respectively, as corresponding to non-problem areas in terms of eutrophication (OSPAR Commission, 2003). The five uppermost stations of the Mondego River estuary and eight stations of the Tejo River estuary would have been classified as Good for nutrients if the OSPAR criteria had been applied. No background concentrations are indicated for the Mira River estuary. These background concentrations, defined as the lowest winter concentrations measured in those estuaries, have to be used with some caution since the oldest monitoring records were registered in the early 80’s, when estuaries were already under considerable human pressure. Bricker et al. (2003) consider that nutrients are not a robust descriptor of eutrophication in estuaries and use the NEAA approach to identify problem and non-problem areas, based on several indicators of pressure, state and response of the aquatic system (Bricker et al., 2003). Based on the NEAA methodology, the Mondego River estuary was identified as a potential problem area by Ferreira et al. (2003), while the Tejo and Mira estuaries were considered non problem areas. Nevertheless, this evaluation was based on data from only the lower Mira River estuary. In spite of the lower level of human pressure in the Mira River basin when compared to the Mondego and Tejo river basins (Table 5.1), this estuary is also subject to nutrient enrichment, at least in upstream areas, because of a low flow, particularly during the dry period.

Biological indices Diversity

Diversity was an attribute considered in both B-IBI and TICOR, but the former considered habitat specific thresholds, while the latter adopted reference values defined for marine communities in Norway. Diversity indices are, in general, good indicators of change in the community structure but their use as an indicator depends on the natural heterogeneity of the communities studied, since diversity is an ambiguous concept and there is no single definition of high and low diversity. When no reference conditions are available, diversity can only be useful for comparing communities from different locations and/or tracking temporal changes a benthic community at a specific site.

The Shannon-Wiener diversity index is very popular among ecologists as a measure of variety and abundance of organisms, since it incorporates a species richness component, but also a species evenness component (Magurran, 2005), while the Margalef index is considered a species richness index (Read et al., 1978). In this study, the Shannon-Wiener and Margalef indices showed highly correlated results, thus their simultaneous use as responsive metrics is

redundant and overemphasizes the diversity component of TICOR, as indicated by the correlation levels of these indices and the EQR values in all estuaries. As pointed out by some authors (Izsák, 2007; Chainho et al., 2007), the use of correlated indices does not contribute additional information, therefore only one of these indices should be used. The Shannon-Wiener is more widely used and has been included in the majority of papers published on comparing different indices to assess ecological status (e.g. Salas et al., 2004; Labrune et al., 2005; Marín-Guirao et al., 2005; Quintino et al., 2006). Thus it is more suitable for comparisons across estuaries.

Redundancy is also not considered by the metric selection criteria used in B-IBI, since all candidate metrics that showed significant differences between degraded and reference locations and were ecologically meaningful were retained in the index (Weisberg et al., 1997;

Van Dolah et al., 1999; Llansó et al., 2002). Paul et al. (2001) developed a multimetric index for the Virginian Province, conceptually similar to the B-IBI, but using stepwise discriminant analysis to select the best subset of metrics for separating between degraded and reference sites. Some problems concerning the use of this statistical method, such as failure to select the best subset of variables of a given size, sampling error capitalization and the use of incorrect number of degrees of freedom have also been identified and Hulberty (1989) suggests a previous elimination of variables with no predictive validity and variables highly correlated with other variables as an alternative for reducing the number of metrics. An alternative approach to minimize redundancy in multimetric indices would be to combine the results of related metrics (e.g. diversity and dominance indices) into a single value prior to combining it with other metrics, to balance the contribution of different attributes of the benthic community. The indication of the logarithmic base and the abundance data type (abundance per replicate, density, etc.) used to calculate diversity indices must be specified, in order to assure their correct use based on specific thresholds, such as those defined in B-IBI and TICOR.

Pollution-indicative and pollution-sensitive taxa

Biotic indices have long included information on the tolerance and sensitiveness of species to pollution. The B-IBI includes pollution-sensitive and pollution-indicative metrics using either abundance or biomass estimates while the TICOR approach includes the abundance of taxa placed into five ecological groups corresponding to different tolerance levels (using the AMBI index). For the B-IBI a two-step procedure was used to determine which species to include in the pollution-sensitive and pollution-indicative metrics (Weisberg et al., 1997). First a candidate list of species for both metrics was developed using the literature and prior knowledge of each candidate species. Second, for the pollution-sensitive metric a

candidate species had to have a higher abundance in reference samples compared to degraded samples while candidate species for the pollution-indicative metric had to have a higher abundance in degraded samples compared to reference samples. The assignation of species by the AMBI index was based on literature addressing the sensitivity/tolerance of taxa to organic enrichment (Borja et al., 2000).

In the present study, the AMBI index seems to overestimate the benthic status in all estuaries, contributing to the overall higher category obtained with TICOR. The tendency for overestimation when using AMBI has been mentioned by other authors. Marín-Guirao et al.

(2005) and Quintino et al. (2006) attributed AMBI’s overestimation of ecological status to the development of the AMBI based on the species’ tolerance to organic pollution, which might make this index less sensitive to other types of pollution such as metal contamination and physical disturbance. The weight given to dominant species has also been indicated as leading to misclassification, since diversity and the number of species are not considered (Labrune et al., 2006). The effect of the dominance of tolerant species is reflected in the three different estuaries studied, as mentioned before by Dauvin et al. (2007). Tolerant species such as Streblospio shrubsolii (Buchanan, 1890) and Corophium spp. (both in EG III) are dominant in the Mira and Mondego River estuaries, therefore constraining the AMBI results towards a general classification of stations into Good status. In contrast, the more balanced distribution of species by ecological groups in the Tejo River estuary gives more weight to relative abundance of species included in other groups. On one hand, the higher species richness in the Tejo River estuary seems to reflect a higher diversity of habitat types, as shown by different combinations of sediment types and salinity classes. On the other hand, the higher number of sensitive species (EG I and EG II) also indicates that these species find favourable conditions for their settlement in the Tejo River estuary, regardless of the overall highest pollution levels. Furthermore, the number of species and their distribution among ecological groups is very similar in the Mondego and Mira estuaries, but density and biomass are considerably higher in the Mondego River estuary, which seems to be an indication of nutrient enrichment.

Both estuaries are under severe natural stress caused by seasonal and daily changes in salinity and freshwater flow and periodically affected by floods and droughts. Extreme events have severe impacts on benthic communities and affect the results obtained when using biotic indices (Chainho et al., 2006; 2007) in the Mondego River estuary, where the number of taxa and their respective abundances decreased significantly after a flood. Droughts also influence the estuarine water quality because freshwater flow is reduced and temperatures increase, lowering dissolved oxygen levels and increasing salinity (Attrill & Power, 2000), which corresponds to what was observed in the Mira River estuary during summer. The ecological process of community succession, widely documented after Pearson & Rosenberg

(1978), can be greatly affected by natural environmental variability (Rakocinski et al., 2000), such as salinity fluctuation in estuarine environments or even be interrupted by disturbance events (Boesch et al., 1976). Ritter et al. (2005) show that a salinity-stressed estuary is in a constant state of early to intermediate succession and there is no climax community but a constant replacement of tolerant species, according to the existing environmental conditions.

The distribution of species by ecological groups in the Mondego and Mira estuaries is consistent with Ritter et al.’s (2005) characterization of a salinity-stressed estuary, i.e.

dominance of tolerant species and low representation of pollution sensitive and pollution indicative species. In contrast, a higher number of pollution sensitive species was identified in the Tejo River estuary when compared to tolerant species, but pollution indicative species assigned to ecological group IV were very representative, giving some evidence of the presence of different successional states.

Multimetric indices

As emphasized by Borja & Dauer (in press), the use of indices to identify impacts of human pressure over ecosystem function requires that these indices are appropriately applied in space and time and generate results that are interpreted in an acceptable manner. In the present study we tested two multimetric approaches developed for different biogeographical regions, including one proposed for assessing the ecological status of Portuguese transitional waters (TICOR). A previous study by Chainho et al. (2007) pointed out some problems related to the applicability of TICOR without prior temporal stratification due to strong seasonal changes in the Mondego estuary. In this study, only summer data was considered and all different salinity habitats were covered in each estuary.

The classifications obtained in the Mondego, Tejo and Mira estuaries using different benthic indices show that the level of agreement between physical-chemical classifications and benthic classifications were not significantly different among indices used. The discrimination between stations above or below Good physical-chemical status level was higher than when three classification levels (Poor/Bad, Moderate and Good/High) were used, suggesting that none of these indices is sensitive to smaller differences in status, as referred to by Quintino et al. (2006). This seems to indicate the need for a geographical adaptation of the metrics and thresholds of both indices, similarly to what has been referred by other authors (e.g. Van Dolah et al., 1999; Salas et al., 2006, Blanchet et al., in press; Borja &

Dauer, in press), since the B-IBI was developed specifically for the Chesapeake Bay and TICOR is composed of indices whose thresholds were also defined for other biogeographic regions.

Additionally, as indicated by the correlation between pollution indicative variables and benthic indices, only diversity indices seem to respond to the stressors measured and only to

nutrient related ones, corroborating the results of Chainho et al. (2007). Nevertheless, in the Mira River estuary there is no apparent response of the indices to stressors and only AMBI and B-IBI identified the only station in a Good status. The low diversity and richness found in this estuary is apparently the reason for the worse classification by B-IBI and TICOR, when compared to the Mondego and Tejo estuaries.

All indices were correlated to each other in the Tejo estuary, while in the Mondego and Mira estuary only some TICOR components were related. Having no evidence that this fact relies on different pressure levels, it seems likely to be related to differences in hydrographical characteristics. The Tejo estuary is one of the largest European estuaries, with a water volume and residence time much higher than the other estuaries studied, acting as a buffer that reduces variations in parameters such as salinity, temperature and sediment composition. In contrast, the Mondego and Mira estuaries register strong variations in environmental conditions, not only across seasons but also daily variations associated with the tidal cycle. In the Mira estuary salinity amplitudes of 25 units were registered between seasons and measurements along the tidal cycle showed variations up to 15 units during a flood period (unpublished data). As pointed out by Ritter et al. (2005), frequent disturbances in environmental conditions, such as salinity changes, prevent the establishment of equilibrium species and the benthic communities remain in a constant state of early to intermediate succession. This seems to explain the strong correlation between all indices tested in the Tejo estuary, similarly to what had been found in the Chesapeake Bay (Borja et al., in press), since communities characterized by high diversity and pollution-sensitive species were identified in some locations, whereas low diversity and pollution-tolerant species are found in other locations.

CONCLUSIONS

The WFD guidance document on ecological status states that the mismatch between the biological and physical-chemical monitoring results may be an indication that the biological methods used are not sensitive to the effects of anthropogenic changes, emphasizing the need to improve biological methods (Ecostat, 2003). The present study showed that hydromorphological differences between estuaries included in the same WFD type (A2) may confound and complicate the classification process, both for physical-chemical and biological elements.

The use of physical-chemical parameters to define reference conditions still lacks consensus among experts and the benthic communities do not always show predictable responses to all types of stressors. Both the B-IBI and the TICOR approach seem to be efficient

in discriminating between locations above or below Good status, as required by the WFD, but were much less efficient in discriminating other quality classes, indicating the need for further adaptation and validation of metrics. The use of diversity can be misleading in estuaries with a natural low diversity, as stressed by Puente et al. (in press), but this problem could be addressed by defining different thresholds for each habitat type, following the approach used in the B-IBI. Metrics related to the tolerance of species to pollution stress also need some adjustments in estuaries with strong natural pressures, since the dominance of tolerant species often overestimates the ecological status, especially in the case of the AMBI.

This could imply the development of separate sets of metrics and thresholds for the Tejo estuary and for the Mondego and Mira estuaries, since benthic communities are under different levels of natural stress.

The future use of indices to classify the benthic status in Portuguese estuaries will require a better understanding of the spatial and temporal patterns of the invertebrate communities. In estuaries with almost no knowledge available on the benthic community patterns, such as the Mira estuary, a greater monitoring effort will be needed before being able to adapt existing benthic indices to an acceptable level of confidence. Ferraro et al.

(1991) concluded that the effect of natural disturbances on the benthos may sometimes be greater than the effect of wastewater discharge and recommend long term studies (≥6 years) to reliably discriminate between them. The monitoring frequency required by the WFD for the benthic fauna (every 3 years) seems consequently inappropriate for a consistent assessment of the benthic status of estuarine systems with similar characteristics to Mondego and Mira

(1991) concluded that the effect of natural disturbances on the benthos may sometimes be greater than the effect of wastewater discharge and recommend long term studies (≥6 years) to reliably discriminate between them. The monitoring frequency required by the WFD for the benthic fauna (every 3 years) seems consequently inappropriate for a consistent assessment of the benthic status of estuarine systems with similar characteristics to Mondego and Mira